摘要
为解决不敏粒子滤波算法对雷达机动目标跟踪实时性差和跟踪起始阶段收敛慢的问题,引入卡尔曼粒子滤波算法。通过坐标转换将实际的极坐标雷达观测数据转换为直角坐标数据,然后用线性最优的卡尔曼滤波器估计粒子状态先验概率密度,最后用非线性最优的粒子滤波器精确估计目标状态后验概率。仿真实验表明,与不敏粒子滤波相比,卡尔曼粒子滤波以牺牲较少精度(减少约6%)的代价,实现机动目标跟踪的实时性(约为前者的1/5),起始阶段收敛性更好。
The unscented particles filter has poor real-time performance and converges slowly in the beginning of radar maneuvering target tracking. The Kalman particle filter was used to solve the problem. Firstly, the radar measurements which were measured under polar coordinates were transformed into data of Cartesian coordinates. Secondly, the prior probabilistic density of the particles was obtained based on the linear optimal Kalman filters. Then, the posterior probabilistic density of targets' state was computed out using the nonlinear optimal particles. Compared with UPF, the time the KPF consumed was only about 1/5 of the UPF in tracking the radar maneuvering targets, and the cost of KPF was lust about 6% of nrecision.Besides,in the beginning of tracking the KPF converged more quickly.
出处
《电光与控制》
北大核心
2012年第1期50-53,共4页
Electronics Optics & Control
关键词
机动目标跟踪
不敏粒子滤波
卡尔曼粒子滤波
坐标转换
实时性
收敛性
maneuvering target tracking
unscented particle filter
Kalman particle filter
coordinate transformation
convergence